/interactive-model-cards

An experimental project that examines whether interactivity can augment people's understanding of machine learning models through model cards

Primary LanguagePythonMIT LicenseMIT

Interactive Model Cards

Paper (Accepted at ACM Facct'22)

This is an experimental project that examines whether interactivity can augment people's understanding of machine learning models. We build upon the model card work initially introduced by Mitchell et al. (2018) and strategically add (and evaluated) interactivity through a human centered design process.

You can interact with the latest version of the demo in your browser here https://huggingface.co/spaces/nazneen/interactive-model-cards

If you would like to programmatically interact with the tool or contribute, please follow the instructions below.

Getting started

cd interactive_model_cards

pipenv install

streamlit run app.py

Citation

@article{2205.02894,
Author = {Anamaria Crisan and Margaret Drouhard and Jesse Vig and Nazneen Rajani},
Title = {Interactive Model Cards: A Human-Centered Approach to Model Documentation},
Year = {2022},
Eprint = {arXiv:2205.02894},
Doi = {10.1145/3531146.3533108},
}